In our Annual Energy Outlook 2025 (AEO2025) Reference case, we project the electricity consumed for commercial computing will increase faster than any other end use in buildings. Computing accounted for an estimated 8% of commercial sector electricity consumption in 2024 and grows to 20% by 2050. Ultimately, more electricity could be consumed by computing than for any other end use in the commercial sector, including lighting, space cooling, and ventilation.
We expect commercial computing growth will outpace computing efficiency improvements which, in the past, have moderated the growth in electricity consumption associated with computers. Commercial computing electricity demand growth is significant enough in our projections to contribute to a reversal in the trend in declining commercial electricity intensity, as measured in kilowatthours consumed per square foot. Many of our assumptions about future energy consumption are based on average energy consumption per unit of commercial floorspace.
Computing includes energy consumption from data center servers, desktop and laptop computers, and monitors in commercial spaces. Data center computing is significantly more energy intensive than computing in general. By 2050, as much as 7% of all U.S. commercial floorspace requires additional energy to meet data center demand across most building types.
Because data centers generate heat and require more air exchange, the increase in data center computing also requires more commercial ventilation and space cooling. These uses are sensitive to assumptions about population migration and the weather. Without computing demand, ventilation and space cooling would still grow but at slower rates.
Much of what we know about how U.S. commercial buildings use energy is informed by our Commercial Buildings Energy Consumption Survey (CBECS). Data center equipment is found in both standalone data center buildings and in on-premises data center rooms that are part of most building types. The latest CBECS shows that data center rooms were most prevalent in healthcare and large office buildings.
In our AEO2025 projections, much of the growth in data centers is in office buildings and in the other building category, which includes standalone data centers. In our Reference case, we make no additional assumptions about revolutionary technologies such as the accelerating expansion of highly energy-intensive artificial intelligence (AI) or, conversely, initiatives that may moderate the energy consumption of data centers generally or AI specifically.
Our projections are benchmarked to sectoral consumption data and forecasts from our State Energy Data System, Monthly Energy Review, and Short-Term Energy Outlook. Since these consumption measures include all commercial sector activity, including energy use for AI, cryptocurrency mining, and data centers, the energy consumption associated with these uses is implicitly represented in our AEO projections.
Principal contributor: Courtney Sourmehi